ivadomed/microscopy-segmentation-monai

Axon and myelin segmentation of SEM images

ivadomed is an integrated framework for medical image analysis with deep learning.

The technical documentation is available here. The more detailed installation instruction is available there

Installation

ivadomed requires Python >= 3.7 and < 3.10 as well as PyTorch == 1.8. We recommend working under a virtual environment, which could be set as follows:

python -m venv ivadomed_env
source ivadomed_env/bin/activate

Install from release (recommended)

Install ivadomed and its requirements from Pypi <https://pypi.org/project/ivadomed/>__:

pip install --upgrade pip
pip install ivadomed

Install from source

Bleeding-edge developments builds are available on the project's master branch on Github. Installation procedure is the following:

git clone https://github.com/neuropoly/ivadomed.git
cd ivadomed
pip install -e .

Contributors

This project results from a collaboration between the NeuroPoly Lab and the Mila. The main funding source is IVADO.

List of contributors

Consult our Wiki(https://github.com/ivadomed/ivadomed/wiki) here for more help

microscopy-segmentation-monai's Issues

Change 2-class model to 3-class model

At the moment, the model has 2 output channels for the axon and myelin masks. This formulation of the problem is not ideal because the classes are not mutually exclusive and the model gives lots of false positives for the background class. This could be mitigated by having a separate output channel for the background and using one-hot encoding for the loss/metrics computations.

The explanation related to this for the ivadomed implementation can be found in this comment

For reference, the current predictions look like this

Change repo name

I'm not sure if the repo name is appropriate. It was suggested to use the standard ivadomed convention for model name, but I was afraid it would cause confusion with our default model (https://github.com/axondeepseg/default-SEM-model). Also, the issue of underscores in the model name format was raised last meeting.